[Corpora-List] Help in Applying Appropriate Statistical Test and Its Interpretation

Stefan Th. Gries stgries at gmail.com
Tue Jun 29 18:36:35 UTC 2010


> It's that corpus data often wildly violates the assumptions underlying the chi-square statistic, making the p-values meaningless.
You're getting no argument from me there ...However, the fact that,
say, the data points are often not independent is a problem for pretty
much all statistical approaches that do not take that into
consideration. So, yes, mixed-effects models would help and I do use
them (although many of their issues are not yet fully solved), but
rule-of-thumb effect sizes don't since, even if they don't come with a
significance test whose assumptions are violated, they may still be
inflated due to interdependencies between data points.

STG
--
Stefan Th. Gries
-----------------------------------------------
University of California, Santa Barbara
http://www.linguistics.ucsb.edu/faculty/stgries
-----------------------------------------------

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora



More information about the Corpora mailing list